Linus Torvalds on Microkernels

I thought of this article because the fundamental issues in the microkernel vs monolithic kernel debate are the same ones that are coming up now in microservices vs a monolithic application debate.

For example RPCs have overhead and increased operational costs. The simplicity of individual services in the micro-* system does not translate into a simple overall system. Nothing about a monolith prevents good modularity.

These statements are all true, and just as applicable whether we are talking kernel design debates 20 years ago or application design debates today.

Now there are use cases for micro-* where they really do make sense. It is important to understand the capabilities of the systems and to know when those cases are. For example neither QNX or Google would be possible without micro-* designs (for different reasons). But if your answer is "always" or "most of the time", I think that you're wrong.